摘要
本文提出一种新的图像显著区域提取方法,分别提取原始图像的亮度、颜色、方向三个特征,并将三个特征的多尺度图像特征合并成一幅总的显著图;其中,在图像颜色特征提取中融入图像的频域特征,简化了算法的复杂度及实现难度,在图像方向特征提取中应用新的特征函数,使得方向特征图更加完善。实验结果显示本文方法相比较Itti算法显著图更为明显且易于实现,在图像目标重定位应用中图像形变少,效果更好。
A novel algorithm for salient region extraction is presented. First, three characteristics graphs of brightness, color, the orientation of the original image are extracted respectively. Then, multi-scale image features of the three characters are combined into one saliency map. Therein the color feature＇ is obtained in frequency domain, which can simplify the algorithm complexity and make it come true easily. In orientation feature extraction, a new characteristic function is used to make the orientation feature map more complete than Gabor function. The experiment results show that the method proposed is easier to be realized and can extract more obvious salient maps than Itti model. In image retargeting we demonstrate that using our saliency prevents distortions in the important regions
出处
《光电工程》
CAS
CSCD
北大核心
2012年第8期18-25,共8页
Opto-Electronic Engineering
基金
国家科技支撑计划项目(2009BAI71B02)
河北省科技支撑项目(11213518D)
河北省教育厅自然科学重点项目(ZD200911)
关键词
显著区域
显著图
特征提取
视觉注意
图像目标重定位
salient regions
saliency map
feature extraction
visual attention
image retargeting
